When a sudden shock to travel demand left no comparable data to plan from, or to check a model’s read of the crisis against, MASS Analytics turned to an earlier disruption as a historical benchmark for both. That gave the airline a defensible recovery plan before fresh data existed to build one.
Sector: Aviation, international air travel · Scope: A European international airline, 10 destinations across 2 major markets, three years of data · Engagement: Multi-year
Always-ON measurement, delivered on the MassTer platform.
The challenge
A sudden, sector-wide shock to travel demand hit an international airline based in Europe, the kind of disruption with no precedent anywhere in its own historical data. Marketing spend and ticket sales had become disconnected. A traditional flat-equation model could not separate the brand-building effect of central campaigns from the demand-conversion effect of local market activations, nor could it link media investment to ticket sales through the awareness and consideration stages that drive the long booking cycle behind every ticket purchase. Operating across ten destinations in two major markets, the airline needed to plan its recovery immediately, long before enough post-shock data existed to build a reliable model from scratch. That plan also had to hold up to scrutiny, since nothing in the airline’s own history looked like what it was living through.
The previous model could not isolate what mattered most
The airline’s previous flat-equation model measured total marketing spend against ticket sales without separating central brand-building campaigns from local promotional activity. It also never traced media through the awareness and consideration stages that lead to a booked ticket. Out of Home advertising had quietly become saturated, yet the previous model still treated it as if it were as effective as ever. When the shock hit, the model had no way to confirm whether its crisis-period output reflected a genuine change in marketing effectiveness. It might just as easily have broken down under conditions it had never been tested against.
Four blind spots the previous model could not see
- The model measured central brand campaigns and local promotional activity as one blended effect. There was no way to tell how much of any result came from long-lasting brand investment versus short-lived local activity.
- The model never traced media investment through the awareness and consideration stages that lead to a booked ticket. As a result, campaigns that built the funnel without an immediate sale looked like they had no effect at all.
- The team still planned Out of Home advertising as a high-return channel long after it had quietly become saturated. Each additional dollar was generating only a fraction of the return available elsewhere.
- When an unprecedented shock to travel demand hit, the model had no internal benchmark to confirm whether its crisis-period output reflected a genuine change in marketing effectiveness. It had no way to rule out a simple breakdown under conditions it had never faced.
How the rebuild closed each gap
Each gap above was invisible in the previous model’s output. The rebuild addressed all four with an explicit model architecture and a historical benchmark, summarised in Table 1.
| Model Component | Approach | What It Measures | Finding |
|---|---|---|---|
| Brand and local media | Separated central brand-building and local promotional variables, each with its own carryover | How long each type of media’s effect lasts after the campaign ends | Central branding carried a significantly longer effect than local activity, which was more immediate but shorter-lived |
| Full booking funnel | Nested three-stage equation system: awareness, consideration, ticket sales | How media builds through the stages that precede a booked ticket | Revealed which campaigns were building funnel stages that ultimately drove ticket sales, across all ten destinations |
| Crisis response | Historical-proxy technique: an earlier disruption to travel demand used as a benchmark | Whether the model’s crisis-period output reflected genuine change or a breakdown under unprecedented conditions | Confirmed the model’s crisis decomposition was consistent with a known historical analogue, supporting its credibility when no other benchmark existed |
An explicit variable or technique addressed each element of the previous model’s blind spot, rather than leaving it to be absorbed into an unexplained residual.
The solution
MASS Analytics built a nested log-linear model at destination level across all ten markets, structured as a three-stage equation system. Awareness was modeled as an outcome of upper-funnel brand media, and consideration as an outcome of awareness and mid-funnel local activation. Ticket sales, in turn, were modeled as an outcome of consideration and lower-funnel performance channels. The model also incorporated external data on airline capacity, competitive fares, traveler mobility, and competitor activity. That way, demand shifts caused by factors outside the airline’s own media were never mistakenly credited to, or blamed on, its marketing.
Central and local media were measured on two different clocks
Separating central brand-building media from local promotional activity revealed two very different response patterns. Central branding carried a significantly longer effect, continuing to influence bookings well after each campaign ended. Local market actions, by contrast, drove a more immediate but shorter-lived response. Measuring both horizons in a single framework gave the airline a defensible basis to protect its brand investment while sharpening efficiency at the local level. That is exactly the kind of call that is hardest to make with confidence when budgets are under pressure. Out of Home advertising, meanwhile, had become highly saturated, generating diminishing returns from every additional dollar it received.
A historical proxy did double duty: planning and validation
The shock itself created a second problem: the crisis period looked like nothing else in the model’s own historical data. As a result, the standard checks a model normally relies on could not tell a genuine modeling issue apart from the legitimate impact of an unprecedented event. MASS Analytics addressed this with a historical-proxy technique. The team searched the airline’s data for an earlier disruption to travel demand with a comparable shape, then compared the model’s estimated crisis trajectory against that earlier disruption’s observed pattern. The comparison did not produce a precise calibration. It did give the team a directional reference that supported the model’s credibility when no other external benchmark existed. That let the airline plan its recovery on a defensible evidence base, well before enough fresh data existed to build a model in the conventional way.
Results and Impact
Guided by the model, the airline protected its brand investment and reallocated spend away from a saturated channel and into local markets. That shift delivered a 17% increase in Marketing ROI from a more precisely deployed budget.
Two time horizons, measured in one model
The model confirmed that central branding carried a significantly longer effect than local promotional activity, continuing to influence bookings well after each campaign ended. Local market actions, by contrast, drove a more immediate but shorter-lived response. Measuring both horizons side by side gave the airline a defensible basis to protect its brand investment while sharpening efficiency at the local level. That beat treating the two as a single undifferentiated media effect.
Budget moved from a saturated channel into local markets
Out of Home advertising had become highly saturated, generating diminishing returns from every additional dollar it received. The airline reallocated spend away from that channel and into local markets, lifting local media budgets by 25%. It kept investing in brand search throughout. The restructured mix delivered a 17% increase in Marketing ROI, sharpening short-term efficiency across all ten destinations in both major markets. It did this without giving up the long-term brand effect that only central campaigns carried.
Marketing ROI improvement from the restructured mix
Local market budget increase, funded by cutting saturated Out of Home
Destinations modeled
Time horizons measured side by side, brand and performance
Related case studies
- Unified Marketing Measurement: how a major retailer built one measurement framework by fusing experiments with MMM.
- Media Mix Optimization: turning model outputs directly into a reallocated budget plan.
- Sales Drivers Analysis: uncovering the true drivers behind a brand’s sales performance.
See what Marketing Mix Modeling could uncover in your own media mix.
Talk to our team about a model built around your channels, your markets, and your data.
